Dear Davorin, please send us the L-D graph comparing analysis and experiment. There are many possible explanations and without seeing the L-D and the model, we could only guess...
Is it not possible you are using Arc Length, therefore not reaching the full force level while loading, because AL reduces the load to improve convergence, and the reduction is lower during unloading? When a certain load level is to be reached, please use Newton-Raphson and not Arc Length.

Thank you for your answer. The differences at initial cycles are negliable. Problem is at larger force amplitudes (monotonically increasing cyclic load) when , for example i reach 160,0 kN with step of 10,0 kN, i'am trying to bring the load with same no. of steps back to zero, so i can assign horizontal load from another direction.
In the experiment we pushed the specimen first from one side than unloaded, than pushed from another side, after also unloaded.
Probelm in my analysis is i think that when during the unloading phase i actually assign the force in opposite direction to unlaod the specimen, but for the force level of 160,0 kN in my case, the resistance is decrased due to damaging when loading. Because of that my specimen is pushed through initial position in unloading phase while it should stop at initial position. The specimen is weaker when unloading so the specimen cannot be placed in initial position. I think this residual displacement is not real, beacuse of unloading, since in experiment we actually release the pressure i.e. we did not apply force in opposite direction. This effect in num. analysis becomes more pronounced as the force increases.
Problem is as i understand that you must assign the force in opposite direction to cancel the effect of the one in first direction, but the stiffness than in this directions is not the same, so the "residual displacement" is not real as in my eperiments. How to do this unlaoding or is there any way to avoid it?

Regards,

Davorin

I am not sure if I understand what is your problem - you can either return the displacement to zero (displacement control), or the force (force control). Of course you have to do the same way in the analysis as in the experiment.
If the calculated residual displacements are larger than in the experiment, you should (re)check the material properties and the boundary conditions. Does the Load-Displacement response up to the first turning point (max. load in the first cycle) in the analysis follow the experimental measurement, or are there significant differences already at that time?

If you need displacement control, but at the same time you wish to check some results when passing through the initial position, you have to find the corresponding step using your displacement monitor, and the postprocess that step (or maybe just output the monitors of interest for that step).